Machine Learning: Classification Techniques

Each feature represents one pixel's intensity, ranging from 0 (white) to 255 (black).

It involves splitting the training set into three folds and training the model three times, each time holding out a different fold for evaluation.

A confusion matrix counts the number of times instances of one class are classified as another, for all class pairs.

<p>The images are 28 x 28 pixels in size.</p> Signup and view all the answers

<p>An accuracy rate above 95% was achieved on all cross-validation folds.</p> Signup and view all the answers

<p>Precision measures the accuracy of positive predictions made by the classifier.</p> Signup and view all the answers

<p>Precision = 0.84.</p> Signup and view all the answers

<p>Recall signifies the true positive rate, indicating the number of correctly predicted positive instances.</p> Signup and view all the answers

<p>Recall = 0.65.</p> Signup and view all the answers

<p>The F1 score is the harmonic mean of precision and recall.</p> Signup and view all the answers

<p>F1 score = 0.73.</p> Signup and view all the answers

<p>FPR is the ratio of negative instances incorrectly classified as positive, important for understanding classifier errors.</p> Signup and view all the answers

<p>Precision and recall often trade off against one another; increasing one can decrease the other.</p> Signup and view all the answers

<p>True Negative Rate.</p> Signup and view all the answers

<p>It plots the false positive rate against the true positive rate for all possible thresholds.</p> Signup and view all the answers

<p>N x (N-1) / 2.</p> Signup and view all the answers

<p>To reduce false positives and improve the classifier's accuracy.</p> Signup and view all the answers

<p>By measuring the F1 score for each label and then computing the average.</p> Signup and view all the answers

<p>It provides insights into the number of correct and incorrect predictions.</p> Signup and view all the answers

<p>It involves generating new training data by applying transformations to existing data.</p> Signup and view all the answers

<p>It is a simple classifier used for multilabel classification tasks.</p> Signup and view all the answers

A dataset containing 70,000 images of handwritten digits, created with contributions from high school students and US Census Bureau employees.

Each image in the MNIST dataset is made up of 784 individual features, representing the intensity of each pixel in the image. The intensity ranges from 0 (white) to 255 (black).

A technique for evaluating the performance of a machine learning model by splitting the dataset into multiple folds and training the model on different folds while using the remaining fold for evaluation.

A measurement of the performance of a classification model, showing the number of times instances of a particular class are classified as other classes.

A baseline model in machine learning is a simple model that provides a starting point for comparison with more complex models.

The accuracy of positive predictions. It measures how many of the predicted positives are actually positive. Calculated as TP / (TP + FP), where TP is true positives and FP is false positives.

The ability to correctly identify all positive instances. It measures how many of the actual positives are correctly identified. Calculated as TP / (TP + FN), where TP is true positives and FN is false negatives.

A measure that combines both precision and recall, providing a balanced evaluation of the model's performance. Calculated as 2 * (precision * recall) / (precision + recall).

The trade-off between precision and recall, where increasing one might decrease the other. It's about finding the right balance between correctly classifying positive instances and avoiding false positives.

The sensitivity to changes in the decision threshold. It measures how well the model can distinguish between positive and negative instances across different thresholds.

The ratio of negative instances that are incorrectly classified as positive. It's a measure of how often the model predicts a negative instance as positive.

The performance measure that plots the True Positive Rate (TPR) against the False Positive Rate (FPR) for different decision thresholds.

A measure that indicates the percentage of correctly classified instances, both positive and negative. It's calculated as (TP + TN) / (TP + TN + FP + FN), where TP is true positives, TN is true negatives, FP is false positives, and FN is false negatives.

The true negative rate (TNR) is the proportion of actual negatives that are correctly identified as negative by the model. It's also known as specificity.

A precision-recall (PR) curve is a graphical representation of a classification model's performance, plotting precision against recall for different threshold values. It is particularly useful when evaluating models in cases of imbalanced datasets, where the number of instances in each class is highly uneven.

The 'One-vs-Rest' (OvR) strategy is a strategy used in multiclass classification problems, where each class is trained against all other classes. This results in creating a binary classifier for each class, effectively reducing the multiclass problem into several binary classification problems.

The 'One-vs-One' (OvO) strategy is another approach for multiclass classification, where a binary classifier is trained for each pair of classes. For N classes, this approach requires N x (N-1) / 2 binary classifiers.

Data augmentation is a technique used in machine learning to artificially increase the training dataset size by creating new data patterns from existing ones. This helps improve the generalization ability and robustness of the model.

Multilabel classification is a type of classification problem where each instance can belong to multiple classes simultaneously. This allows for more nuanced labeling and recognition of complex patterns.